Add a Superseded revocation disposition; revoke the stranded entry-graph-union-construction receipt

[gunbai-bot]

Summary

• Adds payload-free AcceptanceNodeRemainsSuperseded to AcceptanceRevocationDisposition. The arm names only the revocation consequence (the receipt stays off the active frontier). It does not carry by. SupersededLine already owns terminal standing and by.
• That arm is structurally joined to the referenced node's existing SupersededLine before withholding is admitted (join_remains_superseded_to_declared_line / remains_superseded_revocations_join). A still-live TicketLine, an undeclared node, or two declared rows sharing the same node id is a typed refusal, not first() on an ambiguous match. The join accumulator is sticky: once it refuses, later well-formed events cannot overwrite the refusal.
• Appends an AcceptanceRevoked event for entry-graph-union-construction. The recorded acceptance stays in the log; the successor carries the work.
• Frontier reopen is bound to AcceptanceNodeReopensActiveFrontier only. Replay still removes the live receipt for both arms.
• RoadmapAuthorityProjected carries accepted and withheld from one carrier load. frontier_sets_from_events admits FrontierSetsProjected only after the remains-superseded join, a successful replay, and every live receipt joining the current declaration authority (receipt_node_match_verdict, including LiveReceiptOnSupersededLine). An unreplayable history is FrontierSetsProjectionRefused, never accepted=[] / withheld=[]. Live frontier projections consume both halves from that plan. The old accepted_roadmap_nodes_projection / accepted-id wrappers are deleted.
• Seed JSONL deserializer (src/v1/stage0/src/cli_run/roadmap_acceptance_history_carrier.rs): one new coproduct arm on the existing v1 carrier, plus parse and payload-bearing-refusal tests. This is gunbc.v1_maintenance_standing v1_seed_standing (parse the .dag coproduct), not a new seed surface.

Why this is one PR

Adding the arm without using it would be a dangling declaration. Using a reopen revocation for a superseded node would assert that the node rejoins the active frontier, which is false.

#11303 did not cause this. A live receipt remained after #10631 recut the row to SupersededLine, which has no TicketFields, so receipt_matches_node cannot hold. Teaching receipt_matches_node to skip terminal rows is refused; LiveReceiptOnSupersededLine is the typed cause and is consumed by the frontier projection, not only by a witness.

Discriminating evidence

• witness_superseded_revocation_does_not_reopen_active_frontier — revoked superseded node is not on the active frontier
• witness_reopen_revocation_returns_node_to_active_frontier — a reopen revocation of a still-ticketed node still is
• witness_remains_superseded_revocation_on_ticket_line_refuses / witness_remains_superseded_revocation_on_undeclared_node_refuses / witness_ambiguous_declared_node_refuses_remains_superseded_join
• witness_orphan_reopen_revocation_refuses_frontier_sets — a revocation whose exact prior receipt was never live refuses the frontier projection
• witness_record_after_remains_superseded_on_superseded_line_refuses_frontier_sets — recording again while the node is still SupersededLine cannot reopen via the withheld fold
• witness_live_receipt_on_superseded_line_refuses_with_typed_cause
• CI heal on this PR regenerated ROADMAP.md only: focused hidden-active count 125 → 124

Test plan

• cargo test -p v1-compiler --lib cli_run::roadmap_acceptance_history_carrier::tests
• heal regenerated ROADMAP.md with hidden-active 125 → 124
• required floor / witnesses on the live head (d190bb4b3e29edd5416f5cb30de46e6cab0f55dc at this body rewrite; a body edit does not move HEAD)

[gunbai-bot]

Both findings in the dashboard review artifact (stdout of /api/reviews/65663) were present on the then-current source. They are addressed on this head:

1. withheld_node_ids_from_live_history / _from_carrier_text now return WithheldNodeIdsProjection. LoadRefused is WithheldNodeIdsProjectionRefused, never []. Authority projection from carrier text derives withheld inside the successful load arm from the same events. Discriminating control: witness_withheld_projection_refuses_when_history_load_refuses.
2. startable_nodes_missing_closing_contract_for_accepted and the pair/edge/page/forecast consumers take withheld and exclude it the same way the declared active set does. Discriminating control: witness_withheld_sized_subject_does_not_mint_closing_contract_task (plants a live Sized closing-contract candidate into withheld and asserts it is not minted).

— sent from warm-swift-573

[gunbai-bot]

review 65695 (on f42f4b7) asked the withheld helpers to refuse on LoadRefused instead of mapping to []. That was real on that SHA.

Current head 10e4d390469a709acc4c549ba9c067ef341c27f5 already does it: withheld_node_ids_from_live_history / _from_carrier_text return WithheldNodeIdsProjection, and RoadmapAcceptanceEventHistoryLoadRefused is WithheldNodeIdsProjectionRefused. Empty withheld is no longer a load-failure answer. Discriminating control: witness_withheld_projection_refuses_when_history_load_refuses. Authority projection also carries withheld from the same successful load as accepted, so a caller-supplied plan does not re-read the public carrier.

No further commit for this item.

— sent from warm-swift-573

[gunbai-bot]

review 65737 (on 66fbee) asked withheld to ride on RoadmapAuthorityProjected instead of a second withheld_node_ids_from_live_history() behind a caller-supplied plan. That was real on that SHA.

Current head 10e4d390469a709acc4c549ba9c067ef341c27f5 already does it:

• RoadmapAuthorityProjected carries accepted and withheld
• roadmap_authority_projection_from_carrier_text fills both from one loaded event list (withheld is not discarded)
• roadmap_schedule_forecast_over, roadmap_daily_workspace_page_for_projection_observed, and active_section_for bind withheld from the plan; they do not call withheld_node_ids_from_live_history()

Remaining withheld_node_ids_from_live_history() sites are live-only projections that are not handed a caller plan. No further commit for this item.

— sent from warm-swift-573

[gunbai-bot]

review 65819 asked for two source repairs. Both are on this head:

1. The five live frontier projections no longer pair accepted_roadmap_node_ids_projection with withheld_node_ids_from_live_history. They consume frontier_sets_from_live_history, which produces both halves from one carrier load (the same fold roadmap_authority_projection now uses).
2. receipt_node_match_verdict uses node_line_is_superseded instead of a second match on rn.line.

The seed JSONL carrier growth is the coproduct parse realization; not treated as a blocking item in that review.

— sent from warm-swift-573

[gunbai-bot]

review 65879 asked for two remaining pre-migration roots. Both are on this head:

1. withheld_node_ids_from_carrier_text now consumes frontier_sets_from_events after one load, the same admission as withheld_node_ids_from_live_history. withheld_node_ids_from_events is deleted.
2. finish_line_progress_markup consumes frontier_sets_from_live_history. accepted_roadmap_node_ids_projection / accepted_roadmap_nodes_projection / the carrier-text accepted-id projection are the same fold, so a live receipt that cannot join a declared node refuses instead of silently dropping.

— sent from warm-swift-573

[gunbai-bot]

review 65892: accepted_roadmap_nodes_projection and the accepted-id wrappers are deleted. Tests and the carrier-text refuse chain now consume FrontierSetsProjected / roadmap_authority_projection_from_carrier_text. frontier_sets_from_events is the remaining production root for the paired accepted/withheld answer.

— sent from warm-swift-573